More accurate erf, Bessel functions Re: [Maxima] More accurate gamma

>>>>> "Raymond" == Raymond Toy <toy@rtp.ericsson.se> writes:

    Raymond> While playing around with some elliptic integrals, I needed to
    Raymond> evaluate some gamma functions.  The numerical gamma functions included
    Raymond> with maxima seem to have about 5-6 digits of accuracy.

As pointed to me by willib@unk.edu, the Bessel functions aren't very
accurate either.  I looked around and see that erf isn't that good
either.

I have replacements for these, if anyone is interested.  They're Lisp
versions of the Fortran code in TOMS 715, converted via f2cl.  I've
tested these and they're accurate to about within 2-3 bits for the
most part, according to the tests included in TOMS 715.

Perhaps they'll get added someday, or someone will hook up Pari to
maxima again.

Ray